Hi everyone,

I have to questions; my task is to try to predict future stock return by using last months returns. def. as
Code:
return=(Index/ L.Index) - 1
I tried using
Code:
reg return L.return
aswell as
Code:
arima return, arima(1,0,0)
Which of them fits better, or is there any others?

As the second part of my task is to try to predict future returns using out-of-sample method. I have monthly data from 187101 to 201712, and i would like to try dividing this period 50/50 and seeing if the first part can predict second part. How do i go around to do it

Code:
* Example generated by -dataex-. To install: ssc install dataex
dataex yyyymm Index return, count(30)
clear
input long yyyymm double(Index return)
187101 4.44  .01242389
187102  4.5  .02019115
187103 4.61  .02964644
187104 4.74  .03405454
187105 4.86  .02848361
187106 4.82 -.01542547
187107 4.73 -.00863808
187108 4.79  .01915629
187109 4.84  .01136751
187110 4.59 -.05886838
187111 4.64  .03742423
187112 4.74   .0257288
187201 4.86  .02951672
187202 4.88  .00101633
187203 5.04  .04072984
187204 5.18  .03142682
187205 5.18 -.00213872
187206 5.13 -.00795212
187207  5.1 -.00034236
187208 5.04 -.00761387
187209 4.95 -.01335902
187210 4.97  .01832136
187211 4.95 -.00212152
187212 5.07  .03346823
187301 5.11  .01082923
187302 5.15  .01575765
187303 5.11 -.01307993
187304 5.04 -.00878622
187305 5.05  .01496919
187306 4.98 -.01535528
end